Journal of Hydrology (2008) 355, 148 163
available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/jhydrol
The impact of climate change on the water resources
of Hindukush Karakorum Himalaya region under
different glacier coverage scenarios
a, a b
*
M. Akhtar , N. Ahmad , M.J. Booij
a
Institute of Geology, University of the Punjab, Lahore, Pakistan
b
Department of Water Engineering and Management, University of Twente, Enschede, The Netherlands
Received 20 April 2007; received in revised form 20 February 2008; accepted 16 March 2008
KEYWORDS
Summary Thispaperpresentsestimatesofwaterresourceschangesinthreeriverbasinsin
Regional climate
the Hindukush Karakorum Himalaya (HKH) region associated with climate change. The
model;
present climate (1961 1990) and future climate SRES A2 scenario (2071 2100) are simu-
Delta change
lated by the PRECIS Regional Climate Model at a spatial resolution of 25 · 25 km. Two
approach;
HBV models (i.e. HBV-Met and HBV-PRECIS) are designed to quantify the future discharge.
HBV model;
HBV-Met is calibrated and validated with inputs from observed meteorological data while
Climate change;
HBV-PRECISiscalibratedandvalidatedwithinputsfromPRECISRCMsimulationsforthecur-
Glacier coverage;
rent climate.Thefutureprecipitationandtemperature seriesareconstructedthroughthe
Hindukush Karakorum
deltachange approach inHBV-Met, while in HBV-PRECIS future precipitation andtempera-
 Himalaya region
ture series from PRECIS RCM are directly used. The future discharge is simulated for three
stages of glacier coverage: 100% glaciers, 50% glaciers and 0% glaciers. Generally temper-
atureandprecipitationshowsanincreasetowardstheendof21stcentury.Theefficiencies
of HBV-Met during calibration and validation are higher compared to the HBV-PRECIS effi-
ciencies. In a changed climate, discharge will generally increase in both models for 100%
and50%glacierscenarios.Forthe0%glacierscenario,HBV-Metpredictsadrasticdecrease
inwaterresources(upto94%)incontrasttoHBV-PRECISwhichshowsonlyadecreaseupto
15%. Huge outliers in annual maximum discharge simulated through HBV-Met indicate that
hydrological conditions are not predicted perfectly through the delta change downscaling
approach. The results for HBV-Met simply confirm that the quality of observed data in this
regionispoor.TheHBV-PRECISmodelresultsareindicativeofthehigherriskoffloodprob-
lems under climate change. The climate change signals in all three river basins are similar
however, there aredifferences inthe evaluated future water resources estimated through
HBV-Met, whereas in HBV-PRECIS the changes in water resources are similar. This shows
that the transfer of climate change signals into hydrological changes is more consistent
*
Corresponding author. Tel.: [phone].
E-mail address: [e-mail] (M. Akhtar).
0022-1694/$ - see front matter ª 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.jhydrol.2008.03.015
The impact of climate change on the water resources of Hindukush Karakorum Himalaya region 149
inHBV-PRECISthaninHBV-Met.Oneofthereasonsofthepoorerresultsofthedeltachange
approach isthat inthis approach the frequency of rainy days isnot changed and day today
variabilityintemperatureisnotcorrectlytransferred.Howevermoreresearchisneededto
evaluate the uncertainties in both downscaling approaches. Moreover, the dynamical
downscaling approach needs to be tested with other RCMs and preferably to other river
basins as well.
ª 2008 Elsevier B.V. All rights reserved.
approach (Hay et al., 2002). In the delta change approach,
Introduction
an expected mean temperature change is added to the ob-
served temperature record to obtain a future temperature
Pakistan s economy is agro-based and highly dependent on
time series. Precipitation is usually multiplied by a fraction.
the large scale Indus irrigation system (Snow and Ice Hydrol-
However, the adjustment of precipitation in this way is not
ogy Project, 1990). Water issues hold a unique place in Paki-
ideal,astheresultsdepend onthewayin whichthe monthly
stan s policymaking history. It has generated significant
factors are applied to the daily rainfall series (Reynard
heated debate and controversy for a very long time as illus-
et al., 2001). Another way to estimate the future water re-
trated by issues such as the construction of the Kalabagh
sources is by using RCM outputs directly to force the hydro-
Dam and the unequal distribution of water resources among
logical model (Hay et al., 2002; Graham et al., 2007;
the provinces (Ghazanfar, 2007). Impacts of climate change
Leander and Buishand, 2007). In these studies some bias
and climate variability on the water resources are likely to
corrections are made in the RCM outputs before using them
affect irrigated agriculture and installed power capacity.
into hydrological models calibrated and validated with ob-
Changes in flow magnitudes are likely to raise tensions
served meteorological data. The direct use of RCM output
among the provinces, in particular with the downstream
has the potential advantage that more complex changes in
areas (Sindh province), with regard to reduced water flows
the probability density functions of the input variables of
in the dry season and higher flows and resulting flood prob-
hydrological models are taken into account. Akhtar et al.
lems during the wet season. Therefore, in Pakistan future
(2008) used the RCM outputs directly in the hydrological
water resources estimation under climate change is impor-
model and estimated the parameters of the hydrological
tantfor planning and operation ofhydrological installations.
model for each source of input, i.e. outputs from different
To investigate the impact of climate change on future
RCMs, separately.
water resources a hydrological model can be driven with
One field of application of hydrological models is the cre-
the output from a general circulation model (GCM) (Watson
ation ofrunoff scenarios fordifferent climate and glaciation
et al., 1996). However, the spatial resolution of GCMs
conditions. However, glacier storage is not handled well by
(about 250 km) might be too coarse for hydrological model-
current conceptual or physically-based hydrological models.
ing at the basin scale. One way to bridge this scale gap is
Hence, holistic approaches to study and model glacier stor-
through statistical downscaling (e.g. Wilby et al., 1999). In
age are of major importance to fully integrate glaciers into
many hydrological studies (Bergström et al., 2001; Pilling
the hydrological balance to be used for water resources
and Jones, 2002; Guo et al., 2002; Arnell, 2003; Booij,
and river flow predictions at all time scales (Jansson et al.,
2005) statistical downscaling of different GCMs has been
2003). During the 20th century, most of the world s glaciers
used totranslate the assumed climate change intohydrolog-
have shrunk (Paul et al., 2004; WGMS, 2002; Haeberli et al.,
ical response. An alternative approach is dynamical down-
1999)andforawarmingrateof0.04 K a 1,withoutincreases
scaling (e.g. Hay et al., 2002; Hay and Clark, 2003), in
in precipitation, few glaciers would survive until 2100. On
which a regional climate model (RCM) uses GCM output as
the other hand, if the warming rate is limited to 0.01 K a 1
initialandlateralboundaryconditionsoveraregionofinter-
withanincreaseinprecipitationof10%perdegreewarming,
est. The high horizontal resolution of a RCM (about 25
it is predicted that overall loss would be restricted to 10
50 km) is moreappropriate for resolving the small-scale fea-
20% of the 1990 volume (Oerlemans et al., 1998). The gla-
tures of topography and land use, that have a major influ-
ciers of the Greater Himalaya are also retreating (Mastny,
ence on climatological variables such as precipitation in
2000), although Hewitt (1998) reports the widespread
climate models. Moreover, the high resolution of the RCM
expansion of the larger glaciers in the central Karakorum,
is ideal to capture the variability of precipitation as input
accompanied by an exceptional number of glacier surges.
to hydrological models (Gutowski et al., 2003). Recent
The aim of our study is to examine the impact of climate
applications of this approach are presented by Kay et al.
change on the future water resources of three river basins
(2006a, b) and Leander and Buishand (2007).
of the Hindukush Karakorum Himalaya (HKH) region under
To estimate the impact of climate change on river dis-
different glacier coverage scenarios. To achieve this we
charges different scenarios of the future meteorological
make use of the output of the RCM PRECIS nested within
conditions (e.g. temperature and precipitation) are used
the GCM HadAM3P as input into the HBV hydrological model
as input to a hydrological model of a river basin to calculate
to estimate the discharge of the three river basins in the
the corresponding discharges. Changes in downscaled tem-
present and future climate. The GCM uses the SRES A2
perature and precipitation series can be applied to observed
greenhouse gas emission scenario for the simulation of the
temperature and precipitation series by simple transforma-
future climate. The study area is described in section
tionrules.Wewill refertothisapproach asthedeltachange
150 M. Akhtar et al.
 Description of study area . The climatological inputs and some features of the study basins and Fig. 1 shows the
HBV hydrological model are briefly described in section location of the three river basins. These three river basins
 Methodology . The results of the PRECIS RCM present and are situated in the high mountainous HKH region with many
futuresimulations andimpactsofclimate changeonthe dis- peaks exceeding 7000 m and contain a large area of peren-
charge are presented in section  Results and discussion . Fi- nial snow and ice. The HKH region is dominated by large
nally, the conclusions and recommendations are given in glaciers and there is a fivefold to tenfold increase in pre-
section  Conclusions and recommendations . cipitation from glacier termini ( 2500 m) to accumulation
zones above 4800 m. Maximum precipitation occurs be-
tween 5000 and 6000 m (Hewitt, 1993). Most glaciers are
Description of study area
nourished mainly by avalanche snow. Westerly circulations
and cyclonic storms contribute two third of high altitude
Three river basins are selected for analysis: Hunza river
snowfall (Hewitt et al., 1989), while one third derives from
basin, Gilgit river basin and Astore river basin. Table 1 lists
summer snowfall mainly due to monsoon circulation
(Wake, 1989). A huge loss of ice mass and glacier reces-
sions are observed in almost all Karakorum glaciers for
Table 1 Characteristics of study area most of the 20th century until the mid 1990s. Since then
there has been thickening and advances in many glaciers
River basins
but confined to the highest watersheds of the central
Hunza Gilgit Astore
Karakorum (Hewitt, 2005). In spite of surge type behavior
of some glaciers in the HKH region (Diolaiuti et al., 2003),
Gauging station Dainyor Gilgit Doyian
some others (e.g. the Baltoro glacier) are stable during the
Latitude 35° 560 35° 560 35° 330
last 100 years (Mayer et al., 2006) and glaciers located in
Longitude 74° 230 74° 180 74° 420
valleys are declining. A shift to a positive mass balance
Elevation of gauging station (m) 1450 1430 1583
may be taking place, in accordance with weather-station
Drainage area (km2) 13925 12800 3750
records and gauging stations that show reduced runoff
Glacier covered area (km2) 4688 915 612
from the most heavily glacierized Hunza basin (Fowler
% Glacier covered area 34 7 16
and Archer, 2006; Archer and Fowler, 2004). However,
Mean elevation (m) 4472 3740 3921
the suddenness of the changes in glaciers and their con-
% Area above 5000 m 35.8 2.9 2.8
finement to the highest watersheds suggests that thermal
No. of meteorological stations
and hydrological thresholds being crossed that trigger
Precipitation  2 1
down slope redistribution of ice by normal as well as surg-
Temperature  2 1
ing flow, with or without mass balance changes (Hewitt,
No. of PRECIS grid points 12 10 4
2007).
Figure 1 Location of three river basins.
The impact of climate change on the water resources of Hindukush Karakorum Himalaya region 151
Seasonal snow melt and melting of glacial ice are both andRoulin, 1998; Middelkoop etal.,2001). In thisapproach,
large contributors in the discharge of the selected three riv- the observed climate time series are adapted with esti-
ers. In most winters, 80 90% of the area becomes snow cov- mated monthly climate changes from the PRECIS RCM. The
ered (Snow and Ice Hydrology Project, 1990). Climatic observational database used for the delta change approach
variables are strongly influenced by altitude. The HKH re- covers the period 1981 1996. The future daily temperature
gion receives a total annual rainfall of between 200 and ðTf;dailyÞand daily precipitatioPf;dailyÞtime series are con-
500 mm, but these amounts are derived from valley-based structed by Eqs. (1) and (2), respectively
stations and not representative for elevated zones. High-
Tf;dailyźTo;dailyþðTf;monthly Tp;monthlyÞð1Þ
altitude precipitation estimates derived from accumulation
pits runoff above 4000 m range from 1000 mm to more than
Pf;monthly
3000 mm. These estimates depend on the site and time of Pf;dailyźPo;daily ð2Þ
Pp;monthly
investigation, as well as on the method applied (Winger
et al., 2005). where To;daily is the observed daily temperature, Po;daily is the
observed daily precipitation, Tf;monthly is the mean monthly
PRECIS simulated future temperature, Tp;monthly is the mean
Methodology monthly PRECIS simulated present temperature, Pf;monthly is
the mean monthly PRECIS simulated future precipitation
and Pp;monthly is the mean monthly PRECIS simulated present
Climatological input
precipitation.
Depending on the source of input data, i.e. observed
Observed data
Daily observed meteorological data from the Gilgit and As- meteorologicaldataandPRECISRCMsimulationsforthecur-
rentclimate,twoHBVmodels(seeSection HBVhydrological
tore meteorological stations are selected for the Gilgit
model )arecalibratedandarehereafter referredtoas HBV-
and Astore river basins. There is no meteorological station
MetandHBV-PRECIS,respectively.Theclimatechangesignal
in the Hunza river basin, therefore neighboring Skardu
from the PRECIS RCM is transferred to HBV-Met through the
meteorological station is used for calibration and validation
of HBV. The observed discharge data for the three river ba- delta change approach whereas in HBV-PRECIS the future
simulated temperature and precipitation series are used
sins are available at the outlets of the basins. The length of
directly. The effect of climate change on river discharge is
the records in the three river basins is not the same and
there are some missing years in the discharge data. There- simulated for the current glacier extent (100% glacier sce-
nario) and for two stages of deglacierisation, i.e. after an
fore, in some cases the calibration and validation periods
areal reduction by 50% (50% glacier scenario) and after com-
in the three river basins are not same (see Table 3 in section
plete melting (0% glacier scenario).
 Calibration and validation of HBV model ).
Regional climate model outputs HBV hydrological model
TheRCMusedinthisstudyisPRECISdevelopedbytheHadley
Centre of the UK Meteorological Office. The PRECIS RCM is For river discharge simulation, the hydrological model HBV
basedontheatmosphericcomponentoftheHadCM3climate of the Swedish Meteorological and Hydrological Institute
model (Gordon et al., 2000) and is extensively described in (SMHI) is used (Bergström, 1995; Lindström et al., 1997).
Jones et al. (2004). The atmospheric dynamics module of Using inputs from RCMs this model has reproduced the dis-
PRECISisahydrostaticversionofthefullprimitiveequations charge fairly well for e.g. the Suir river in Ireland (Wang
and uses a regular longitude latitude grid in the horizontal et al., 2006). HBV has been widely used in Europe and other
and a hybrid vertical coordinate. For this study, the PRECIS parts of the world in climate change studies (Liden and Har-
modeldomain(UpperIndusbasin)hasbeensetupwithahor- lin, 2000; Bergström et al., 2001; Menzel and Bürger, 2002;
izontalresolutionof25 · 25 km,ascomparedtoAkhtaretal. Booij, 2005; Menzel et al., 2006). This model is a semi-dis-
(2008) who used a horizontal resolution of 50 · 50 km. The tributed, conceptual hydrological model using sub-basins
domain is roughly stretched over the latitude 26° 39°N and as the primary hydrological units. It takes into account
longitude 67° 85°E. The HadAM3P global data set is used area-elevation distribution and basic land use categories
to drive the PRECIS model. The horizontal resolution of the (glaciers, forest, open areas and lakes). Sub-basins are con-
HadAM3Pboundarydatais150 kmandforthepresentandfu- sidered in geographically or climatologically heterogeneous
ture climate, it covers the period 1960 1990 and 2070 basins. The model consists of a precipitation routine repre-
2100, respectively (Wilson et al., 2005). For the future cli- senting rainfall and snow, a soil moisture routine determin-
mate, the SRES A2 greenhouse gas emission scenario is se- ing actual evapotranspiration and controlling runoff
lected (Nakicenovic et al., 2000). formation, a quick runoff routine and a base flow routine
The first year in each PRECIS RCM experiment is consid- which together transform excess water from the soil mois-
ered as a spin-up period and these data are not used in ture zone to local runoff, a transformation function and a
any analysis. After post processing of each PRECIS RCM routing routine. A general description of the HBV model is
experiment the time series of temperature and precipita- given in SMHI (2005) and the application of HBV to the
tion are produced for further analysis. HKH region is extensively studied by Akhtar et al. (2008).
As input, the model needs the distribution of the basin
Delta change approach to observed data area by altitude and land use categories, where the glaci-
The delta change approach has been used in many climate ated parts have to be treated as a separate land use class
change impact studies before (see e.g. Arnell, 1998; Gellens and glacier mass balance is determined for each elevation
152 M. Akhtar et al.
zone. For running the daily model, the only required data is performed to assess the sensitivity of the discharge re-
are daily means of temperature and daily total precipitation gime to the parameters. For the three river basins, param-
(potential evapotranspiration is calculated using a simpli- eters GMELT (glacier melting factor), FC (maximum soil
fied version of Thornthwaite s equation with temperature moisture storage), PERC (percolation from upper to lower
as input). Daily discharge is needed for calibration. Param- response box), TT (threshold temperature), DTTM (value
eters of the HBV model are calibrated using a manual cali- added to TT to reach threshold temperature for snowmelt),
bration procedure (SMHI, 2005). In previous HBV studies, and CFMAX (factor for snow melt) are found to be most sen-
much experience has been gained in parameter estimation, sitive and there is a strong interdependence among these
which is used to acquire the range of parameters in our parameters. Therefore, a multivariate sensitivity analysis
study (Uhlenbrook et al., 1999; Krysanova et al., 1999; is performed to calibrate the parameters of the HBV-MET
SMHI, 2005; Booij, 2005). A univariate sensitivity analysis and HBV-PRECIS models for each river basin.
10
(a) Hunza river basin
5
0
-5
-10
-15
-20
-25
-30
JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC
Month
Future temperature Present temperature
20
15
(b) Gilgit river basin
10
5
0
-5
-10
-15
-20
-25
JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC
Month
Future temperature Present temperature
20
(c) Astore river basin
15
10
5
0
-5
-10
-15
-20
JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC
Month
Future temperature Present temperature
Figure 2 Mean annual cycle of temperature over (a) Hunza river basin (b) Gilgit river basin (c) Astore river basin as simulated with
PRECIS for present (1961 1990) and future (2071 2100) day climate (°C).
Temperature (ºC)
Temperature (ºC)
Temperature (ºC)
The impact of climate change on the water resources of Hindukush Karakorum Himalaya region 153
In order to assess the performance of the model, the and Qo is the mean of Qo over the calibration/validation
Nash Sutcliffe efficiency coefficient NS (Nash and Sutcliffe, period. For a favorable model performance, the efficiency
1970) and the relative volume error RE are commonly calcu- NS should be as high as possible (maximum value of 1) and
lated by Eqs. (3) and (4): the RE value should be close to zero.
PiźN
½QsðiÞ QoðiÞŠ2
NSź1 Piź1 ð3Þ
Results and discussion
iźN
½QoðiÞ QoŠ2
iź1
PiźN
Changes of temperature and precipitation during
½QsðiÞ QoðiÞŠ
iź1
REź100 ð4Þ
PiźN 2071 2100
QoðiÞ
iź1
where i isthe time step, N is the total number oftime steps, For three river basins, the mean annual cycles of tempera-
Qs represents simulated discharge, Qo is observed discharge ture and precipitation for the present and future climate
10
(a) Hunza river basin
8
6
4
2
0
JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC
Month
Future precipitation Present precipitation
7
(b) Gilgit river basin
6
5
4
3
2
1
0
JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC
Month
Future precipitation Present precipitation
10
(c) Astore river basin
8
6
4
2
0
JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC
Month
Future precipitation Present precipitation
Figure 3 Mean annual cycle of precipitation over (a) Hunza river basin (b) Gilgit river basin (c) Astore river basin as simulated with
PRECIS for present and future day climate (mm/day).
Precipitation (mm/day)
Precipitation (mm/day)
Precipitation (mm/day)
154 M. Akhtar et al.
simulated with PRECIS are presented in Figs. 2 and 3, models during calibration and validation periods for the
respectively. These indicate a general increase in tempera- three river basins. The average simulated discharge and
ture and precipitation during the period 2071 2100. The average observed discharge is equal during the calibration
warming is uniformly distributed over the three river basins. period for each HBV model and for each river basin and
Bothinthe present andfuture simulatedclimate the highest consequently the relative volume error is zero. General
temperature is reached in the month of July and lowest testing of conceptual models (Rango, 1992) has shown that
temperature is observed duringthe month of January. Table NS values higher than 0.8 are above average for runoff
2 presents the seasonal changes in temperature and precip- modeling in glaciated catchments. Therefore, model effi-
itation in the three river basins with climate change. The ciencies during calibration are satisfactory for the two
annual mean temperature rise by the end of the century is HBV models. The efficiencies of HBV-Met are higher than
up to 4.8 °C. The warming is stronger during the winter sea- those of HBV-PRECIS. Since the calibration period for the
son compared to the summer season. two models are different and the efficiency values highly
PRECIS estimates a rise in annual mean precipitation (up depend on the time period for which the model is run,
to19%)bytheendofthe21stcentury.Theincreaseinprecip- the smaller efficiency of HBV-PRECIS might be due to the
itation is observed in all seasons. Generally, the changes in events that are not captured by the model during that per-
precipitation during the summer season are larger than dur- iod. During the validation period NS values, RE values and
ing the winter season. This is because of the fact that the visual inspection of hydrographs (Figures are not given)
HKHregionreceivesaverysmallamountofprecipitationdur- show that performance of the two HBV models is satisfac-
ingthe summerseason and asmall absolute increase in sum- tory. The values of the performance criteria show that dur-
mer precipitation compared to winter precipitation gives a ing the validation period overall performance of HBV-Met
larger percentual precipitation change during the summer. (e.g. 0.71 < NS < 0.91) is somewhat better compared to
The mean annual precipitation changes in the Hunza (19%), the calibration period (e.g. 0.67 < NS < 0.86), while overall
Gilgit (21%) and Astore (13%) river basins are similar. The performance of HBV-PRECIS during validation (e.g.
general increase in temperature and precipitation is consis- 0.58 < NS < 0.72) is somewhat less compared to the calibra-
tent with the projected increase in temperature and precip- tion period (e.g. 0.74 < NS < 0.82). Although in this study
itation in neighboring areas such as southwest China and we used PRECIS RCM outputs at 25 km resolution (for Upper
northwest India (Yinlong et al., 2006; Kumar et al., 2006). Indus Basin domain) yet on average the efficiency of HBV-
PRECIS is similar to that of Akhtar et al. (2008) achieved
by using PRECIS RCM outputs at 50 km resolution (for South
Calibration and validation of HBV model
Asia domain). This shows that increasing the resolution of
data does not necessarily increase the efficiency of the
Table 3 presents the efficiencies, relative volume error and
hydrological model.
mean observed and simulated discharge for the two HBV
Table 2 Seasonal changes of mean temperature and precipitation under SRES A2 scenario from PRECIS in 2071 2100 over three
river basins relative to 1961 1990
River basins Temperature change (°C) Precipitation change (%)
Annual Winter Summer Annual Winter Summer
Hunza 4.5 4.8 4.2 19 27 10
Gilgit 4.8 4.8 4.8 21 19 24
Astore 4.5 4.7 4.4 13 1 25
(Summer = April September; Winter = October March).
Table 3 Performance of two HBV models during calibration and validation in different river basins
River basin Calibration Validation
Period Qo m3/s Qs m3/s NS RE % Period Qo m3/s Qs m3/s NS RE %
HBV-Met Hunza 1981 1990 306.5 306.5 0.86 0 1990 1996 281.1 275.1 0.91 2
Gilgit 1981 1990 266.8 266.9 0.83 0 1990 1996 295.6 256.7 0.77 13
Astore 1981 1990 133.4 133.4 0.67 0 1990 1996 171.2 152.3 0.71 11
HBV-PRECIS Hunza 1971 1980 370.6 369.8 0.74 0 1981 1990 306.5 364.9 0.72 19
Gilgit 1961 1970 288.7 288.6 0.82 0 1981 1990 266.8 264.8 0.72 0
Astore 1975 1983 120.6 121.6 0.79 0 1983 1990 140.5 109.5 0.58 22
The NS values are for daily model runs.
The impact of climate change on the water resources of Hindukush Karakorum Himalaya region 155
this case has to be regarded as a hypothetical one because
Future annual discharge cycle
future 100% glacier extent is not realistic with climate
change. If the glacierised area is reduced by 50%, snowmelt
Fig. 4 shows the mean annual discharge cycle simulated by
stillbeginsonemonthearlieranddischargeisincreaseddur-
HBV-Met for the present and future climate for three stages
of glacier coverage: 100% glaciers, 50% glaciers and 0% gla- ing March, April and May in the Astore river basin, while it is
ciers. The amplitude of the annual discharge cycle is in- decreased in the Hunza and Gilgit river basins in this period.
The discharge under 50% glacier scenario is also increased in
creased in a changed climate under the 100% glacier
scenario. Snow melting starts one month earlier and dis- September and October in the Hunza river basin. The qual-
itative changes in monthly runoff for the 0% glacier scenario
charge rises towards its peak in summer (August). However,
2000
(a) Hunza river basin
1500
1000
500
0
JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC
Month
Future Discharge-100% glaciation Future Discharge-50% glaciation
Future Discharge-0% glaciation Present simulated discharge
1400
1200 (b) Gilgit river basin
1000
800
600
400
200
0
JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC
Month
Future Discharge-100% glaciation Future Discharge-50% glaciation
Future Discharge-0% glaciation Present simulated discharge
500
(c) Astore river basin
400
300
200
100
0
JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC
Month
Future Discharge-100% glaciation Future Discharge-50% glaciation
Future Discharge-0% glaciation Present simulated discharge
Figure4 Annual discharge cyclesimulatedbyHBV-Metforthe presentclimate andfutureclimate forthree stagesofglaciation for
three river basins.
3
Discharge (m /s)
3
Discharge (m /s)
3
Discharge (m /s)
156 M. Akhtar et al.
are more controversial. The discharge is reduced drastically is increased in a changed climate under the 100% glacier
in all three river basins. HBV-Met shows that the major con- scenario as well. Snowmelt starts one month earlier, i.e.
tribution in these river basins is because of glacial melt, it starts in March. There is an increase in river discharge
although the exact contribution is not known. throughout the year and in all river basins. The highest peak
Fig. 5 shows the mean annual discharge cycle simulated is observed in July. For the 50% glacier scenario the dis-
by HBV-PRECIS for the present and future climate for three charge is increased during March July while during Au-
stages of glacier coverage: 100% glaciers, 50% glaciers and gust October the shape of the hydrograph is generally the
0% glaciers. The amplitude of the seasonal discharge cycle same as in the present climate. After complete reduction
2000
(a) Hunza river basin
1500
1000
500
0
JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC
Month
Future Discharge-100% glaciation Future Discharge-50% glaciation
Future Discharge-0% glaciation Present simulated discharge
1400
1200
(b) Gilgit river basin
1000
800
600
400
200
0
JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC
Month
Future Discharge-100% glaciation Future Discharge-50% glaciation
Future Discharge-0% glaciation Present simulated discharge
500
(c) Astore river basin
400
300
200
100
0
JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC
Month
Future Discharge-100% glaciation Future Discharge-50% glaciation
Future Discharge-0% glaciation Present simulated discharge
Figure 5 Annual discharge cycle simulated by HBV-PRECIS for the present climate and future climate for three stages of glaciation
for three river basins.
3
Discharge (m /s)
3
Discharge (m /s)
3
Discharge (m /s)
The impact of climate change on the water resources of Hindukush Karakorum Himalaya region 157
all glaciation scenarios are also more similar in HBV-PRECIS
Table 4 Mean relative change in future discharge (2071
compared to HBV-Met. Although these basins have similar
2100) in a changed climate relative to the present discharge
geology and hydrology, there is a chance that due to po-
(1961 1990) for three glaciations stages and for three river
tential biases in PRECIS RCM simulated temperature and
basins
precipitation series for the current climate, the parame-
Model River Mean change in discharge (%)
ters of HBV-PRECIS are adjusted in such a way that the im-
basin
pacts on water resources are similar. However,
100% glacier 50% glacier 0% glacier
transferring the climate change signals to the hydrological
HBV-Met Hunza 88 10 65
model through the direct use of RCM simulations preserves
Gilgit 70 12 94
future extremes which is an advantage over the delta
Astore 48 12 72
change approach (Graham et al., 2007).
HBV-PRECIS Hunza 59 21 15
Gilgit 60 24 12
Future discharge peaks
Astore 41 13 15
Extreme value analysis based on the Gumbel extreme value
distribution is carried out to estimate the impact of climate
change on floods for three river basins with two HBV mod-
of the glaciers, snowmelt still starts one month earlier and els and for three glaciation stages. For this, the maximum
there isan increase in discharge during March May. There is discharge per hydrological year is determined from both
a considerable decrease in discharge during July measured and simulated discharge series of three river ba-
September. sins. Fig. 6 shows the extreme value distribution of floods
Table 4 presents the mean relative changes in future derived from observed discharge data and simulated dis-
discharge (2071 2100) in a changed climate relative to charge data from HBV-Met and HBV-PRECIS. Observed dis-
the present discharge (1961 1990) for the three glacia- charge data from the period 1981 1996, simulated
tions stages and for three river basins. There is a big dis- discharge data for HBV-Met from the period 1981 1996
crepancy between the results of changes in discharge and simulated discharge data for HBV-PRECIS from the per-
simulated by HBV-Met and HBV-PRECIS. Under the 100% iod 1961 1990 are used. Since the extreme discharge re-
glacier scenario both models predict an increase in water turn values are influenced by the period of study, it is
resources. However the increase is higher in HBV-Met difficult to compare the observed extreme values with
compared to HBV-PRECIS. Under the 50% glacier scenario HBV-PRECIS simulated extreme values. Moreover, observed
HBV-Met predicts an increase in the discharge in the Hunza and HBV-Met simulated extreme values are based on rela-
river basin, while in the Gilgit and Astore river basins the tively few extreme flood events, which makes the extrapo-
discharge is expected to decrease. HBV-PRECIS predicts lation to large return periods highly susceptible to errors.
an increase in all river basins under the 50% glacier sce- Anyhow, the general trend of present day simulated annual
nario. Without glaciers, HBV-Met predicts a drastic maximum discharge from both HBV models is an underesti-
decrease in the discharge (65 94%) in all river basins, mation at all return levels. The highest differences be-
whereas HBV-PRECIS predicts about a 15% decrease in tween observed and modeled extreme discharges are
the discharge. There is neither forest nor any major lake found in the Astore river basin.
present in the three river basins and glaciers and fields The flood frequency results under climate change for
(area without forest) are considered as the only two land three glacier stages estimated through HBV-PRECIS and
use classes in the hydrological model framework. There- HBV-Met models are presented in Figs. 7 and 8, respec-
fore, the effect of complete melting of glaciers on the tively. In all river basins, HBV-PRECIS shows an increase
hydrological cycle will depend on the degree of glaciation in flood magnitude for all return periods under climate
in the river basins and response of the river basins to cli- change in the 100% and 50% glacier scenarios. The magni-
mate change. For instance looking at the similar patterns tude of flood frequency under climate change in the 0% gla-
of climate change in the three river basins the highly gla- cier scenario is increased in the Hunza and Gilgit river
ciated Hunza river is expected to react more severely com- basins whereas in Astore river basin it is comparable with
pared to the least glaciated Gilgit river basin. However, the current magnitude of floods at least at higher return
HBV-Met shows that more drastic changes are expected periods. The change in peak discharge at 20-year return
in the Gilgit river basin compared to the Hunza river basin. level in the 100% glacier scenario is 68%, 36% and 34% in
This is may be because of the inaccurate transfer of cli- the Hunza, Gilgit and Astore river basins, respectively.
mate change signals through the delta change approach. These results are consistent with the study of Milly et al.
The decrease in discharge predicted by HBV-PRECIS is con- (2002) who found an overall increase in flood peaks during
sistent with the complete melting of glaciers, because the the twentieth century and this trend is expected to con-
net annual ice losses due to wastage of glaciers represents tinue in the future. HBV-Met predicts an increase in flood
between 12% and 15% of the annual water yield from melt- magnitude for all return period under climate change in
ing ice (Hewitt et al., 1989). The transfer of climate the Hunza and Gilgit river basins for the 100% glacier sce-
change signals into hydrological changes seems to be more nario whereas for the 50% and 0% glacier scenarios, the
consistent in HBV-PRECIS. The temperature and precipita- magnitude of peak discharges is decreased. The flood fre-
tion changes are almost similar in all three river basins. quency is increased in the Astore river basin for the 100%
The resulting changes in water resources conditions under glacier scenario for all return periods and for 50% glacier
158 M. Akhtar et al.
Return period (years)
1
10 20 50
2
5
3000
HBV-Met
(a) Hunza River Basin
2500
HBV-PRECIS
Observed
2000
1500
1000
500
0
-2 -101234
Reduced Gumbel variate
Return period (years)
1 20 50
5
2 10
3000
HBV-Met
(b) Gilgit River Basin
2500
HBV-PRECIS
Observed
2000
1500
1000
500
0
-2 -1 0 1 2 3 4
Reduced Gumbel variate
Return period (years)
1 5 10
2 20 50
2000
HBV-Met
(c) Astore River Basin
HBV-PRECIS
1500
Observed
1000
500
0
-2 -101234
Reduced Gumbel variate
Figure 6 Observed, HBV-Met simulated and HBV-PRECIS simulated annual maximum discharge as a function of return period for
three river basins.
scenario at least at higher return periods. For this river flood peaks at 20-year return level are increased (43%)
basin the magnitude of the peak discharge is decreased for the 50% glacier scenario and are decreased (19%) for
for the 0% glacier scenario. The change in peak discharge the 0% glacier scenario.
at 20-year return level in the 100% glacier scenario is The characteristics of future annual maximum discharge
54%, 32% and 73% in the Hunza, Gilgit and Astore river ba- values are given in Table 5. There are huge outliers in HBV-
sins, respectively. For the 50% and 0% glacier scenarios the Met simulated future annual maximum discharge values in
flood peaks at 20-year return level decrease in the Hunza all river basins (not shown in the figures). The outliers
(10% and 54%, respectively) and Gilgit (27% and 83%, are also present in HBV-PRECIS simulated future annual
respectively) river basins. In the Astore river basin the maximum discharge values in the Astore river basin. Some
3
Discharge (m /s)
3
Discharge (m /s)
3
Discharge (m /s)
The impact of climate change on the water resources of Hindukush Karakorum Himalaya region 159
Return period (years)
1 10 20 50
2
5
4000
Future-100% glaciation Future-50% glaciation
3500
Present simulated
Future-0% glaciation
3000
2500
2000
1500
1000
(a) Hunza River Basin
500
0
-2 -1 0 1234
Reduced Gumbel variate
Return period (years)
1
5
2 50
20
10
3000
Future-100% glaciation Future-50% glaciation
2500
Present simulated
Future-0% glaciation
2000
1500
1000
500
(b) Gilgit River Basin
0
-2 -1 01234
Reduced Gumbel variate
Return period (years)
1 5 10
2 20 50
1500
Future-100% glaciation Future-50% glaciation
Present simulated
Future-0% glaciation
1000
500
(c) Astore River Basin
0
-2 -1 01234
Reduced Gumbel variate
Figure 7 HBV-PRECIS simulated annual maximum discharge as a function of return period for current and changed climate for
three glacier stages for three river basins.
of these outliers in both HBV-Met and HBV-PRECIS may be tion. Unfortunately, sufficient precipitation stations are
because of the high variability of runoff due to the small not available to assess the areally averaged basin scale pre-
size of the river basin. The outliers in HBV-Met are ex- cipitation in a right way. Consequently, observed precipita-
plained by the fact that in each river basin only one mete- tion shows too much variability and extreme behavior.
orological station is used for temperature and precipitation Parameters are estimated under too variable and extreme
input into HBV-Met. Observed precipitation is considered as conditions. For example in Hunza river basin at Skardu
areally averaged precipitation but actually point precipita- meteorological station there are three heavy rainfall spells
3
Discharge (m /s)
3
Discharge (m /s)
3
Discharge (m /s)
160 M. Akhtar et al.
Return period (years)
1 10 20 50
2
5
4000
Future-100% glaciation Future-50% glaciation
3500
(a) Hunza River Basin
Present simulated
Future-0% glaciation
3000
2500
2000
1500
1000
500
0
-2 -101234
Reduced Gumbel variate
Return period (years)
1
25 50
10
20
3000
Future-100% glaciation Future-50% glaciation
(b) Gilgit River Basin
2500
Present simulated
Future-0% glaciation
2000
1500
1000
500
0
-2 -101234
Reduced Gumbel variate
Return period (years)
1 5 10
2 20 50
1500
Future-100% glaciation Future-50% glaciation
(c) Astore River Basin
Present simulated
Future-0% glaciation
1000
500
0
-2 -101234
Reduced Gumbel variate
Figure 8 HBV-Met simulated annual maximum discharge as a function of return period for current and changed climate for three
glacier stages for three river basins.
in the month of October 1987 (average rainfall is 37.0 mm The modeled changes in flood frequency under climate
in October, 1987 while the climate normal for October is change are just estimations that are based on simulations
6.4 mm). When we use climate change scenarios derived using input data from only one RCM run, using one emission
from PRECIS (in October there is an increase in precipita- scenario and one single GCM for the boundary data. Other
tion of 57%) HBV-Met gives extremely high peaks in October GCMs could result in quite different flood frequency predic-
1987 (an increase in mean discharge of 289% in October tions. Despite all uncertainties, the behavior of peak dis-
1987). Therefore, the quality of input data used in HBV- charges predicted by the two HBV models supports the
Met seems to be too poor to simulate extreme discharge direct use of RCM output as input to hydrological models in
behavior. this area.
3
Discharge (m /s)
3
Discharge (m /s)
3
Discharge (m /s)
The impact of climate change on the water resources of Hindukush Karakorum Himalaya region 161
Table 5 Characteristics of future annual maximum discharge simulated by two HBV models in a changed climate for the three
glaciations stages and for three river basins
Model River basin Glaciation scenario
Means (m3/s) Standard deviation (m3/s)
100% 50% 0% 100% 50% 0%
HBV-PRECIS Hunza 2052.5 1738.0 1391.9 272.2 294.9 266.5
Gilgit 1540.5 1351.6 1186.0 261.1 288.1 276.6
Astore 549.8 (613.3) 473.9 (536.2) 394.7 (451.7) 321.3 (465.3) 324.1 (462.2) 305.6 (429.2)
HBV-Met Hunza 2123.7 (2837.4) 1140.3 (1887.7) 249.9 (915.5) 265.1 (2775.9) 169.4 (2899.3) 203.5 (2585.1)
Gilgit 1400.2 (1430.5) 688.3 (709.9) 42.9 (53.5) 120.0 (165.0) 65.9 (104.8) 47.3 (61.3)
Astore 566.2 (591.7) 377.8 (401.0) 207.9 (231.3) 99.5 (137.6) 121.2 (147.3) 82.5 (120.5)
The values in parentheses are for future annual maximum discharge with outliers.
hydrological model may be an alternative in areas where
Conclusions and recommendations
the quality of observed data is poor. The direct use of
RCM outputs (HBV-PRECIS model) has shown that the magni-
The PRECIS RCM present climate and future SRES A2 climate
tude of annual maximum flood peaks is likely to increase in
scenarios presented in this paper include detailed regional
the future. Hence, overall results are indicative of a higher
information (25 · 25 km) and is very important for climate
risk of flood problems under climate change. The modeled
impact assessment in various sectors. This paper includes
only the basic aspects of simulated present and future cli- changes in future discharge and changes in flood frequency
mate (i.e. future changes in temperature and precipita- under climate change are not conclusive because more re-
search is needed to evaluate the uncertainties in this ap-
tion). Generally, temperature and precipitation shows an
increase towards the end of the 21st century spread monot- proach. Moreover, this technique needs to be tested with
onously over the three river basins. There are several uncer- other RCMs and preferably to river basins in other parts of
the world as well.
tainty sources in the PRECIS RCM simulations which are not
discussed here. However, we plan to evaluate and quantify
these uncertainties in this region in PRECIS RCM simulations
in our future work.
Acknowledgements
In a changed climate, HBV does not calculate the new
glacier area size automatically. To bridge this deficiency,
Part of this study was completed at ICTP, Trieste, Italy,
we have used three glacier coverage scenarios as applied
which provided funding for two months fellowship for this
by Hagg et al. (2007) while modelling the hydrological re-
work. The Higher Education Commission of Pakistan has also
sponse to climate change in glacierized Central Asian
provided financial funding. The authors would like to thank
catchments. However, future glacier extent may be pre-
the PRECIS team at Hadley Centre, Met Office, UK, for their
dicted separately by using a simple hypsographic modelling
comments and suggestions during the PRECIS simulations.
approach (e.g. Paul et al. 2007). The use of such a pre-
The boundary data of different GCMs have been kindly sup-
dicted future glacier extent in HBV would give a more real-
plied by David Hein on behalf of Hadley Centre, Met Office,
istic hydrological change. To quantify the future water
UK. The daily river discharge data and meteorological data
resources, the delta change approach is used for HBV-
have been taken from WAPDA, Pakistan and Pakistan Mete-
Met and direct use of PRECIS RCM data is done for HBV-
orological Department, respectively. The authors also thank
PRECIS. There are differences in the results of both ap-
the HBV support team at SMHI for their useful comments
proaches. In a changed climate, the discharge will gener-
and suggestions during the study.
ally increase both in HBV-PRECIS and HBV-Met in the
100% glacier scenario (up to 60% and 88%, respectively)
and in the 50% glacier scenario (up to 24% and 10%, respec- References
tively). For the 0% glacier scenario under climate change, a
drastic decrease in water resources (up to 94%) in HBV-Met Akhtar, M., Ahmad, N., Booij, M.J., 2008. Use of regional climate
model simulations as input for hydrological models for the
is present, whereas HBV-PRECIS shows a decrease up to
Hindukush Karakorum Himalaya region. Hydrology and Earth
15%.
System Sciences Discussions 5, 865 902.
There are huge outliers in annual maximum discharge
Archer, D.R., Fowler, H.J., 2004. Spatial and temporal variations in
simulated with HBV-Met. This shows that the prediction of
precipitation in the upper Indus Basin, global teleconnections
hydrological conditions through the delta change approach
and hydrological implications. Hydrology and Earth System
is not ideal in the HKH region. HBV-PRECIS provides results
Science 8, 47 61.
on hydrological changes that are more consistent with
Arnell, N.W., 1998. Climate change and water resources in Britain.
RCM changes. This shows that the climate change signals
Climatic Change 39, 83 110.
in HBV-PRECIS are transmitted more realistically than in
Arnell, N.W., 2003. Relative effects of multi-decadal climatic
HBV-Met. Therefore, the direct use of RCM outputs in a variability and changes in the mean and variability of climate
162 M. Akhtar et al.
due to global warming: future streamflow in Britain. Journal of Hewitt, K., 2007. Tributary glacier surges: an exceptional concen-
Hydrology 270, 195 213. tration at Panmah Glacier, Karakoram Himalaya. Journal of
Bergström, S., 1995. The HBV model. Chapter 13 of Computer Glaciology 53, 181 188.
models of watershed hydrology. Water Resource Publications, Jansson, P., Hock, R., Schneider, T., 2003. The concept of glacier
443 476. storage: a review. Journal of Hydrology 282, 116 129.
Bergström, S., Carlsson, B., Gardelin, M., Lindström, G., Petters- Jones, R.G., Noguer, M., Hassell, D.C., Hudson, D., Wilson, S.S.,
son, A., Rummukainen, M., 2001. Climate change impacts on the Jenkins, G.J., Mitchell, J.F.B., 2004. Generating High Resolution
runoff in Sweden-assessments by global climate models, dynam- Climate Change Scenarios Using PRECIS. Met Office Hadley
ical downscaling and hydrological modelling. Climate Research Centre, Exeter, UK, p. 40.
16, 101 112. Kay, A.L., Jones, R.G., Reynard, N.S., 2006a. RCM rainfall for UK
Booij, M.J., 2005. Impact of climate change on river flooding flood frequency estimation. I. Method and validation. Journal of
assessed with different spatial model resolutions. Journal of hydrology 318, 151 162.
Hydrology 303, 176 198. Kay, A.L., Jones, R.G., Reynard, N.S., 2006b. RCM rainfall for UK
Diolaiuti, G., Pecci, M., Smiraglia, C., 2003. Liligo Glacier, flood frequency estimation. II. Climate change results. Journal
Karakoram, Pakistan: a reconstruction of the recent history of of hydrology 318, 163 172.
a surge-type glacier. Annals of Glaciology 36, 168 172. Krysanova, V., Bronstert, A., Muller-Wohlfeil, D.I., 1999. Modelling
Fowler, H.J., Archer, D.R., 2006. Conflicting signals of climate river discharge for large drainage basins: from lumped to
change in the upper Indus Basin. Journal of Climate 19, 4276 distributed approach. Hydrological Sciences Journal 44, 313
4293. 331.
Ghazanfar, M., submitted for publication. Kalabagh Dam and the Kumar, R.K., Sahai, A.K., Kumar, K.K., Patwardhan, S.K., Mishra,
Water Debate in Pakistan. Lahore Journal of Policy Studies. P.K., Revadekar, J.V., Kamala, K., Pant, G.B., 2006. High-
Gellens, D., Roulin, E., 1998. Streamflow response of Belgian resolution climate change scenarios for India for the 21st
catchments to IPCC climate change scenarios. Journal of century. Current Science 90, 334 345.
Hydrology 210, 242 258. Leander, R., Buishand, T.A., 2007. Resampling of regional climate
Gordon, C.C., Cooper, C., Senior, C.A., Banks, H., Gregory, J.M., model output for the simulation of extreme river flows. Journal
Mitchell, J.F.B., Wood, R.A., 2000. The simulation of SST, sea of Hydrology 332, 487 496.
ice extents and ocean heat transport in a version of the Hadley Liden, R., Harlin, J., 2000. Analysis of conceptual rainfall-runoff
centre coupled model without flux adjustment. Climate Dynam- modelling performance in different climates. Journal of Hydrol-
ics 16, 147 168. ogy 238, 231 247.
Graham, L.P., Andreasson, J., Carlsson, B., 2007. Assessing climate Lindström, G., Johansson, B., Persson,M., Gardelin, M., Bergström,
change impacts on hydrology from an ensemble of regional S.,1997.DevelopmentandtestofthedistributedHBV-96model.
climate models, model scales and linking methods-a case study Journal of Hydrology 201, 272 288.
on the Lule River basin. Climatic Change 81, 293 307. Mastny, L., 2000. Melting of Earth s Ice Cover Reaches New High.
Guo,S.,Wang,J.,Xiong,L.,Yin,A.,Li,D.,2002.Amacro-scaleand WorldWatch Institute,Washington,DC, semi-distributedmonthlywaterbalancemodeltopredictclimate alerts/000306.html>.
changeimpacts in China. Journal of Hydrology 268, 1 15. Mayer, C., Lambrecht, A., Belo, M., Smiraglia, C., Diolaiuti, G.,
Gutowski, W.J., Decker, S.G., Donavon, R.A., Pan, Z., Arritt, R.W., 2006. Glaciological characteristics of the ablation zone of
Takle, E.S., 2003. Temporal-spatial scales of observed and Baltoro glacier, Karakoram, Pakistan. Annals of Glaciology 43,
simulatedprecipitation in centralUS climate. Journalof Climate 123 131.
16, 3841 3847. Menzel, L., Bürger, G., 2002. Climate change scenarios and runoff
Haeberli, W., Frauenfelder, R., Hoelzle, M., Maisch, M., 1999. On response in the Mulde catchment (Southern Elbe, Germany).
rates and acceleration trends of global glacier mass changes. Journal of Hydrology 267, 53 64.
Geografiska Annaler 81, 585 591. Menzel, L., Thieken, A.H., Schwandt, D., Bürger, G., 2006. Impact
Hagg, W., Braun, L.N., Kuhn, M., Nesgaard, T.I., 2007. Modelling of of climate change on the regional Hydrology-Scenarios-Based
hydrological response to climate change in glacierized Central modelling studies in the German Rhine Catchment. Natural
Asian catchments. Journal of Hydrology 332, 40 53. Hazards 38, 45 61.
Hay, L.E., Clark, M.P., 2003. Use of statistically and dynamically Middelkoop, H., Daamen, K., Gellens, D., Grabs, W., Kwadijk,
downscaled atmospheric model output for hydrologic simula- J.C.J., Lang, H., Parmet, B.W.A.H., Schädler, B., Schulla, J.,
tions in three mountainous basins in the western United States. Wilke, K., 2001. Impact of climate change on hydrological
Journal of Hydrology 282, 56 75. regimes and water resources management in the Rhine Basin.
Hay, L.E., Clark, M.P., Wilby, R.L., Gutowski, W.J., Leavesley, Climatic Change 49, 105 128.
G.H., Pan, Z., Arritt, R.W., Takle, E.S., 2002. Use of regional Milly, P.D.C., Wetherald, R.T., Dunne, K.A., Delworth, T.L., 2002.
climate model output for hydrological simulations. Journal of Increasing risk of great floods in a changing climate. Nature 415,
Hydrometeorology 3, 571 590. 514 517.
Hewitt, K., Wake, C.P., Young, G.J., David, C., 1989. Hydrological Nash, J.E., Sutcliffe, J.V., 1970. River flow forecasting through
investigations at Biafo Glacier, Karakoram Himalaya: an impor- conceptual models 1: a discussion of principles. Journal of
tant source ofwater for the Indus River. Annals of Glaciology13, Hydrology 10, 282 290.
103 108. Nakicenovic, N., Alcamo, J., Davis, G., de Vries, B., Fenhann, J.,
Hewitt, K., 1993. The altitudinal distribution of Karakoram geo- Gaffin, S., Gregory, K., Grübler, A., Jung, T.Y., Kram, T., La
morphic processes and depositional environments. In: Shroder, Rovere, E.L., Michaelis, L., Mori, S., Morita, T., Pepper, W.,
Jr., J.F., (Ed.), Himalaya to the Sea: geology, geomorphology Pitcher, H., Price, L., Riahi, K., Roehrl, A., Rogner, H-H.,
and the Quaternary. NY, Routledge, pp. 159 183. Sankovski, A., Schlesinger, M., Shukla, P., Smith, S., Swart, R.,
Hewitt, K., 1998. Glaciers receive a surge of attention in the van Rooijen, R., Victor, N., Dadi, Z., 2000. IPCC Special Report
Karakoram Himalaya. EOS, Transactions. American Geophysical on Emission Scenarios, Cambridge University Press, Cambridge,
Union 79, 104 105. United Kingdom and New York, NY, USA.
Hewitt, K., 2005. The Karakoram anomaly? Glacier expansion and Oerlemans, J., Anderson, B., Hubbard, A., Huybrechts, Ph.,
the elevation effect, Karakoram Himalaya. Mountain Research Johannesson, T., Knap, W.H., Schmeits, M., Stroeven, A.P.,
and Development 25, 332 340. van de Wal, R.S.W., Wallinga, J., Zuo, Z., 1998. Modelling the
The impact of climate change on the water resources of Hindukush Karakorum Himalaya region 163
response of glaciers to climate warming. Climate Dynamics 14, Wake, C.P., 1989. Glaciochemical investigations as a tool for
267 274. determining the spatial and seasonal variation of snow accumu-
Paul, F., Maisch, M., Rothenbuhler, C., Hoelzle, M., Haeberli, W., lation in the Central Karakorum, Northern Pakistan. Annals of
2007.Calculationandvisualisationoffutureglacierextentinthe Glaciology 13, 279 284.
Swiss Alps by means of hypsographic modelling. Global and Wang, S., McGrath, R., Semmler, T., Sweeney, C., Nolan, P., 2006.
Planetary Change 55, 343 357. The impact of the climate change on discharge of Suir river
Paul,F.,Kaab,A.,Maisch,M.,Kellenberger, T.,Haeberli,W.,2004. catchment (Ireland) under different climate scenarios. Natural
Rapid disintegration of Alpine glaciers observed with satellite Hazards and Earth System Sciences 6, 387 395.
data. Geophysical Research Letters 31, L21402. doi:10.1029/ Watson, R.T., Zinyowera, M.C., Moss, R.H., 1996. Climate Change
2004GL02081. 1995: Impacts, Adaptations, and Mitigations of climate change.
Pilling, C.G., Jones, J.A.A., 2002. The impact of future climate Cambridge University Press, 889 pp.
change on seasonal discharge, hydrological processes and WGMS, 2002. World Glacier Monitoring Service, Glacier Mass
extreme flows in the upper Wye experimental catchment, mid- Balance Data 2000/01, .
Wales. Hydrological Processes 16, 1201 1213. Wilby, R.L., Hay, L.E., Leavesley, G.H., 1999. A comparison of
Rango, A., 1992. Worldwide testing of the snowmelt runoff model downscaled and raw GCM output: implications for climate
with applications for predicting the effects of climate change. change scenarios in the San Juan River Basin, Colorado. Journal
Nordic Hydrology 23, 155 172. of Hydrology 225, 67 91.
Reynard, N.S., Prudhomme, C., Crooks, S.M., 2001. The flood Wilson, S., Hassell, D., Hein, D., Jones, R., Taylor, R., 2005.
characteristics of large UK rivers: potential effects of changing Installing and using the Hadley Centre regional climate model-
climate and land use. Climatic Change 48, 343 359. ling system, PRECIS (version 1.4), Met Office Hadley Centre,
SMHI., 2005. Integrated hydrological modelling system (IHMS), Exeter, UK.
Manual version 5.6, SMHI, Norrkoping. Winger, M., Gumpert, M., Yamout, H., 2005. Karakorum Hinduk-
Snow and Ice Hydrology Project, 1990. Snow and Ice Hydrology ush western Himalaya: assessing high-altitude water resources.
Project, Upper Indus Basin, Overall. Report, WAPDA- IDRC- Hydrological Processes 19, 2329 2338.
Wilfrid Laurier University, p. 179. Yinlong, X., Xiaoying, H., Yong, Z., Wantao, L., Erda, L., 2006.
Uhlenbrook, S., Seibert, J., Leibundgut, C., Rodhe, A., 1999. Statistical Analysis of climate change scenarios over China in the
Prediction uncertainty of conceptual rainfall-runoff models 21st century. Advances in Climate Change Research 2, 50 53.
caused by problems in identifying model parameters and
structure. Hydrological Sciences Journal 44, 779 797.